ADSKAILab/floora-1.7b

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 25, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

FLOORA-1.7b by Autodesk Research is a 1.7 billion parameter autoregressive causal language model, fine-tuned from Qwen3-1.7B, specifically designed for architectural floor-plan generation. It excels at creating structured multifamily residential layouts from building metadata and massing geometry, outputting a compact, deterministically parseable architectural Domain-Specific Language (DSL). This model is optimized for rapid conceptual design and exploration within the Architecture, Engineering, and Construction (AEC) domain, offering a unique solution for automated floor-plan generation.

Loading preview...

FLOORA-1.7b: Domain-Specific LLM for Architectural Floor-Plan Generation

FLOORA-1.7b, developed by Autodesk Research, is a specialized 1.7 billion parameter language model (fine-tuned from Qwen3-1.7B) focused on generating structured multifamily residential floor plans. Unlike general-purpose LLMs, FLOORA operates on a custom, compact architectural Domain-Specific Language (DSL) that is token-efficient, human-readable, and deterministically parseable, significantly reducing the data size compared to formats like IFC.

Key Capabilities

  • Structured Floor-Plan Generation: Autoregressively generates labeled floor-plan polygons for living units, corridors, and vertical circulation cores from architectural prompts.
  • Domain-Specific Language (DSL): Utilizes a custom DSL for outputs, enabling deterministic parsing, validation, normalization, rendering, and conversion into geometric floor plans.
  • Human-Aligned Design: Incorporates architect-edited supervised fine-tuning and reinforcement learning with learned architect preferences and verifiable geometric/functional rewards.
  • Efficient Parameterization: Although nominally 1.7B parameters, vocabulary resizing reduces its effective parameter count, enhancing efficiency for its specialized task.

Good For

  • Rapid Conceptual Design: Ideal for quick generation and exploration of multifamily residential layouts.
  • Automated Layout Exploration: Supports automated design exploration and architect review workflows.
  • Architectural Design Research: Useful for research in structured AEC generation and domain-specific language models for engineering.
  • Geometry-Aware Design Optimization: Facilitates design optimization through its structured and verifiable outputs.

FLOORA-1.7b is intended for conceptual proposals and design exploration, not as a replacement for professional architectural judgment or for generating construction-ready documents. Its specialization in multifamily residential layouts means performance may vary on highly irregular or out-of-scope geometries.